An error-based video quality assessment method with temporal information

Videos are amongst the most popular online media for Internet users nowadays. Thus, it is of utmost importance that the videos transmitted through the internet or other transmission media to have a minimal data loss and acceptable visual quality. Video quality assessment (VQA) is a useful tool to determine the quality of a video without human intervention. A new VQA method, termed as Error and Temporal Structural Similarity (EaTSS), is proposed in this paper. EaTSS is based on a combination of error signals, weighted Structural Similarity Index (SSIM) and difference of temporal information. The error signals are used to weight the computed SSIM map and subsequently to compute the quality score. This is a better alternative to the usual SSIM index, in which the quality score is computed as the average of the SSIM map. For the temporal information, the second-order time-differential information are used for quality score computation. From the experiments, EaTSS is found to have competitive performance and faster computational speed compared to other existing VQA algorithms.